Bug 164461 – [EFL] Use libgcrypt instead of GnuTLS for CryptoDigest

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit EFL

[Gyuyoung Kim]

As GTK port in r208297, EFL port starts to use libgcrypt instead of GnuTLS as well.

[Gyuyoung Kim]

Created attachment 294019 [details]
Patch

[Olivier Blin]

(In reply to comment #1)
> Created attachment 294019 [details]
> Patch

Looks good.

Once this lands, we can remove the two GnuTLS classes containing the previous implementation, and move the remaining GnuTLS stubs to the gcrypt folder.

[Michael Catanzaro]

Comment on attachment 294019 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=294019&action=review

r+, though I've been having some second thoughts about this myself. Since we'll already use GnuTLS anyway by default, maybe it makes sense to stick with that one crypto library for everything, and have the gcrypt support here just as a fallback for embedded device vendors that want to avoid GnuTLS.

Shrug!

> Source/cmake/OptionsEfl.cmake:124
> +WEBKIT_OPTION_DEFAULT_PORT_VALUE(ENABLE_SUBTLE_CRYPTO PUBLIC ON)

You probably don't want to do this! This feature is incomplete and unstable and buggy and nobody is working to fix that. It's enabled in FeatureList.pm for developers.

[Gyuyoung Kim]

Created attachment 294022 [details]
Patch for landing

[Gyuyoung Kim]

(In reply to comment #3)

> > Source/cmake/OptionsEfl.cmake:124
> > +WEBKIT_OPTION_DEFAULT_PORT_VALUE(ENABLE_SUBTLE_CRYPTO PUBLIC ON)
> 
> You probably don't want to do this! This feature is incomplete and unstable
> and buggy and nobody is working to fix that. It's enabled in FeatureList.pm
> for developers.

 Ok, I restore it to off.

[Gyuyoung Kim]

Created attachment 294023 [details]
Patch for landing

[Michael Catanzaro]

Comment on attachment 294023 [details]
Patch for landing

Remember that we can't commit today while Apple's continuous integration is broken! Got to wait until that's fixed....

[Michael Catanzaro]

Although, this should really be safe, since it's EFL-specific, maybe it's fine.

[Gyuyoung Kim]

Created attachment 294040 [details]
Patch

[Gyuyoung Kim]

(In reply to comment #8)
> Although, this should really be safe, since it's EFL-specific, maybe it's
> fine.

Yes, this patch is really EFL specific.

[Olivier Blin]

(In reply to comment #3)
> Comment on attachment 294019 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=294019&action=review
> 
> r+, though I've been having some second thoughts about this myself. Since
> we'll already use GnuTLS anyway by default, maybe it makes sense to stick
> with that one crypto library for everything, and have the gcrypt support
> here just as a fallback for embedded device vendors that want to avoid
> GnuTLS.
> 
> Shrug!

gcrypt is now used for ClearKey support in EME, see bug 162928.
If you want to go back on this, OpenSSL would actually be a better alternative for embedded devices.

[WebKit Commit Bot]

Comment on attachment 294040 [details]
Patch

Rejecting attachment 294040 [details] from commit-queue.

Failed to run "['/Volumes/Data/EWS/WebKit/Tools/Scripts/webkit-patch', '--status-host=webkit-queues.webkit.org', '--bot-id=webkit-cq-02', 'validate-changelog', '--check-oops', '--non-interactive', 294040, '--port=mac']" exit_code: 1 cwd: /Volumes/Data/EWS/WebKit

ChangeLog entry in ChangeLog contains OOPS!.

Full output: http://webkit-queues.webkit.org/results/2468793

[Gyuyoung Kim]

Created attachment 294109 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 294109 [details]
Patch

Clearing flags on attachment: 294109

Committed r208440: <http://trac.webkit.org/changeset/208440>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.